SemiconductorsBRIEF

Bristol Myers Squibb (BMS) Deploys Second NVIDIA DGX SuperPOD Built on Eight Vera Rubin NVL72 Systems

N
NathanTechnology Editor · Technical Lead
Published · Updated
According to a NVIDIA blog post, Bristol Myers Squibb (BMS) is deploying its second NVIDIA DGX SuperPOD, built on eight DGX Vera Rubin NVL72 systems delivering up to 10x the performance per megawatt of the infrastructure it replaces. The move follows roughly three years running its first SuperPOD, which BMS says had become saturated.

Why BMS Needed to Upgrade Its AI Infrastructure

According to NVIDIA's blog post, Bristol Myers Squibb (BMS) has operated a DGX SuperPOD for about three years, "producing meaningful results" (E5). One concrete output NVIDIA points to: the BMS team used AI to expand its library of CELMoD compounds — molecules engineered to selectively degrade cancer-causing proteins, with applications in blood cancer treatment and beyond (E6).

But that first-generation system has run out of headroom. Erin Davis, BMS's vice president of research business insights and technology, is quoted in NVIDIA's blog post saying: "We're saturated. We're in production with some very large-scale predictions around large molecules. We're building our own foundational models, and that takes a lot of GPUs" (E7).

New Hardware: Eight DGX Vera Rubin NVL72 Systems

To address that saturation, BMS announced it is deploying its second NVIDIA DGX SuperPOD, this one built on eight DGX Vera Rubin NVL72 systems, according to NVIDIA's blog post (E1). Each of the eight rack-scale systems combines NVIDIA Vera CPUs and Rubin GPUs, and NVIDIA reports the cluster delivers up to 10x the performance per megawatt of the infrastructure it replaces (E2).

MetricValueEvidence
New DGX Vera Rubin NVL72 systems deployed8E1
Performance-per-megawatt gain vs. replaced infrastructureup to 10xE2
Time BMS has operated its first DGX SuperPODabout 3 yearsE5
Payal Sheth's expanded SVP role beganJanuary 2026E4

Erin Davis gave the new cluster an informal name of her own: NVIDIA's blog post quotes her calling it the "SuperDuperPOD" (E3).

A Unified Platform Across BMS's Global Sites

According to NVIDIA's blog post, Davis's team is combining the existing DGX SuperPOD and the new DGX Vera Rubin NVL72-powered system into a single environment — described as "a single data plane, accessible from every BMS site globally" (E8). Read together with the saturation problem described by Davis (E7), the unified platform effectively pools the original three-year-old cluster (E5) and the new eight-system deployment (E1) into one global resource rather than two separate pools.

Lowering the Barrier to Use: Mission Control and Natural-Language Access

NVIDIA's blog post states that barriers which made the earlier system hard to reach — "site-specific restrictions left over from past acquisitions" and "the need for deep computational expertise" — are being replaced with AI-native tooling managed through NVIDIA Mission Control (E9). Under this setup, NVIDIA reports that researchers will be able to initiate complex predictions in plain English (E9).

Bio-AI Tooling: The BioNeMo Agent Toolkit

Per NVIDIA's blog post, the unified platform will give BMS researchers access to a unified AI platform that includes the NVIDIA BioNeMo Agent Toolkit for biological AI — used for running predictions, training models, and powering agentic workflows across the full drug discovery pipeline (E10). This toolkit sits alongside the Mission Control interface (E9) as part of the same unified environment described in NVIDIA's blog post (E8).

Organizational Change: A New Leadership Role for Therapeutic Discovery Sciences

NVIDIA's blog post also notes a leadership change tied to this period: Payal Sheth, a scientist who spent her career inside drug discovery labs, took on an expanded role in January as senior vice president of therapeutic discovery sciences at BMS (E4).

What This Means

Taken together, the evidence in NVIDIA's blog post traces a sequence: BMS ran its first DGX SuperPOD for about three years and used it to expand its CELMoD compound library (E5, E6), then hit saturation while building its own foundational models (E7). The response was an eight-system DGX Vera Rubin NVL72 deployment rated at up to 10x the performance per megawatt of what it replaces (E1, E2), merged with the existing cluster into one globally accessible platform (E8) and made easier to use through Mission Control's natural-language access (E9) and the BioNeMo Agent Toolkit (E10). Payal Sheth's expanded therapeutic discovery sciences role, which began in January — before the July deployment announcement — falls within the same period NVIDIA's blog post describes for this infrastructure shift (E4).

📊 Evidence

FAQ

How many new DGX Vera Rubin NVL72 systems is Bristol Myers Squibb (BMS) deploying?

Eight, according to NVIDIA's blog post announcing the deployment.

What performance improvement does the new system offer over BMS's existing infrastructure?

NVIDIA's blog post states the new systems deliver up to 10x the performance per megawatt of the infrastructure they replace.

How long had BMS been running its first DGX SuperPOD before this announcement?

About three years, per NVIDIA's blog post, during which BMS says it produced meaningful results, including AI-driven expansion of its CELMoD compound library.

📎 Sources

  1. blogs.nvidia.com
N
NathanTechnology Editor · Technical Lead

Related

BRIEF

At SIGGRAPH 2026, NVIDIA Announces Cosmos 3 Edge, DGX Station and MotionBricks for Agentic and Physical AI

According to NVIDIA's SIGGRAPH 2026 blog post, the company announced the 4-billion-parameter Cosmos 3 Edge model, DGX Station systems built on the GB300 Grace Blackwell Ultra Desktop Superchip with up to 20 petaflops of FP4 compute, a Synthetic Video Detector microservice reaching 92% accuracy on uncompressed video, and MCP-based AI agent integrations across five creative-software platforms.

Nathan ·
BRIEF

Intel's New Mexico Fab Pushes Packaging Reticle Limit to 8x Today, Targets 12x+ by 2028

According to Intel's newsroom announcement, Intel's Fab 9 advanced packaging site in Rio Rancho, New Mexico has scaled its 'reticle limit' to 8 times the current industry standard, with a target of over 12 times by 2028. Intel says the facility grew from 25 employees in 1980 to 2,700 employees and 500 suppliers today, using Foveros and EMIB/EMIB-T technologies to package AI chips.

Nathan ·